Performance Persistence

"Past performance is no guarantee of future results" is printed on every fund disclosure for a reason grounded in data, not just legal caution. If you rank managers into quintiles by their trailing five-year return and then track what the top quintile does next, the honest finding across decades of mutual fund research is that it barely outperforms the average — and the very worst quintile persists more reliably (in being bad) than the best quintile does in staying good.

Measuring it

The standard test sorts managers into groups by past performance, then measures the correlation between past rank and future rank, or equivalently builds a "winners minus losers" portfolio and checks whether it earns a positive return going forward:

$$r_{persistence} = \bar{r}_{top\,quintile,\,next\,period} - \bar{r}_{bottom\,quintile,\,next\,period}$$

In words: take the funds that ranked best last period, take the funds that ranked worst, and see how much better the former group did than the latter afterward. A number reliably above zero is evidence of persistence; a number near zero says last period's ranking told you nothing.

Worked example

A study tracks 500 equity mutual funds, sorting them into quintiles by trailing 3-year return. One year later, the top quintile from year one averages a return ranked in the 52nd percentile — statistically indistinguishable from the 50th percentile a coin flip would produce. The bottom quintile, however, averages a rank in the 34th percentile — still bad, meaning its members were disproportionately likely to stay in the bottom half. The asymmetry is the real finding: high expense ratios and poor risk control are persistent liabilities, but good stock-picking in one period is not a reliable predictor of good stock-picking in the next.

What this means in practice

Allocators who chase last year's top performer are, on average, chasing noise plus a fee headwind — the fund that just had a hot streak also just attracted new money, which itself tends to drag down future returns through capacity constraints. The more defensible use of a track record is negative screening: persistently bad funds (high fees, high turnover, poor risk-adjusted returns) really do tend to stay bad, so ruling those out is more reliable than chasing the current leaderboard.